Métodos bayesianos para análise de dados de eventos recorrentes considerando uma classe geral de modelos com fragilidade multiplicativa

• Ano de defesa: 2010
• Autor(a) principal: Gouvêa, Graziela Dutra Rocha
• Orientador(a): Scalon, João Domingos
• Banca de defesa: Muniz, Joel Augusto, Fogo, José Carlos, Cirillo, Marcelo Ângelo, Tomazella, Vera Lúcia Damasceno
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: UNIVERSIDADE FEDERAL DE LAVRAS
• Programa de Pós-Graduação: DEX - Departamento de Ciências Exatas
• País: BRASIL
• Palavras-chave em Português: Inferência bayesiana; Teste de normalidade; Hemodiálise; Bayesian inference; Survival analysis
• Área do conhecimento CNPq: Estatística
• Link de acesso: https://repositorio.ufla.br/handle/1/4301

In some situations, we may observe that the event of interest occurs repeatedly in the same individual, such as when a patient diagnosed with cancer tends to relapse over time or when a person is repeatedly readmitted in a hospital. In this situation, it is more reasonable to suppose the existence of a relationship between times until the occurrence of events. In order to examine the association between survival times, a class of models has been studied. These are called frailty models and consider a random effect in the Cox model. Penã and Hollander (2004) proposed a new class of models, more general and flexible, that simultaneously incorporates the effect of covariates, the impact on the unit of accumulating event occurrences, the effect of latent or unobserved variables which, for each unit, endow correlation among the inter-event times, as well as the effect of performed interventions after each event occurrence.With respect to Bayesian approach, the frailty models have already been explored by many authors. By the other side, the general class of models for recurrent events has not been studied under Bayesian approach. Then, the aim of this thesis was to explore Bayesian methods for the general class of frailty models for recurrent event data analysis and also study, under this focus, the multiplicative frailty model and the Cox model. The models were compared using the selection methods AIC (Akaike information criterion) and BIC (Bayesian information criterion). The proposed methodology was applied to data of chronic renal failure from Santa Casa de Misericórdia in Lavras, MG, which was collected only for this study. The effect of covariates was studied and the individual frailties for the patients in treatment of hemodialysis were estimated. We conclude that the general class of models is important because consider, beside of effects in other models, the effect of cumulating times of recurrent events in the individuals. The application to the dataset showed the importance of both frailty terms and the parameter which measures the effect of cumulating times of recurrent events. It was showed that the model purposed by Peña and Hollander can be useful to analyse recurrent times data of hospitalization to patients in treatment of hemodialysis. In this case, it was checked that the effect of cumulating cumulating recurrent events is an important feature to predict a new hospitalization.